15
Vent: I was training a model on my own data for six months before a friend pointed out the flaw.
I kept feeding it thousands of my own photos to try and get better image recognition. My friend asked to see my setup last week and just said, 'Your labels are a mess, man.' I was using different words for the same thing, like 'car' and 'auto.' That's why the accuracy never got above 65%. Has anyone else fixed a huge labeling mistake and seen a real jump in performance?
2 comments
Log in to join the discussion
Log In2 Comments
taraj113d ago
Your labels are a mess" is such a brutal but helpful catch.
3
river_rivera453d ago
Man, that's rough. I had the same thing with a plant ID model where I used "leaf" and "foliage" randomly. How did you clean it up? I made a strict word list and used a script to find and replace all the old labels, and my accuracy shot up like 30 points. @taraj11 is right, a brutal catch saves so much time later.
3